Transformer fault diagnosis early warning method based on big data

Abstract

The invention belongs to the technical field of transformer fault diagnosis, and discloses a transformer fault diagnosis early warning method based on big data, and the method comprises the steps: collecting transformer operation parameter data and environment parameter data; preprocessing the operation parameter data and the environment parameter data of the transformer to obtain an operation parameter data set and an environment parameter data set; performing feature extraction on the operation parameter data set and the environment parameter data set to obtain an operation feature data set and an environment feature data set; performing weighted fusion on the operation feature data set and the environment feature data set to obtain a comprehensive feature data set; constructing a transformer fault early warning model, and inputting the comprehensive feature data set into the transformer fault early warning model to obtain a predicted transformer operation state; comparing the predicted running state of the transformer with a preset running state threshold value of the transformer, and judging whether the running state of the transformer is normal or not; potential faults of the transformer can be identified more accurately, so that the misdiagnosis rate is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of transformer fault diagnosis, and more specifically, to a transformer fault diagnosis and early warning method based on big data. Background Art

[0002] With the continuous development and expansion of the power system, the operating status of the transformer, as the core equipment for power transmission and distribution, is directly related to the safety and stability of the entire power system.

[0003] The patent application with the publication number CN109871660B discloses a method for early warning of thermal faults in main transformers and a method for locating faults. The present invention relates to the field of methods for handling thermal faults in main transformers. A research on transformer safe operation characteristic monitoring and active early warning technology based on multi-source data fusion is proposed, and the research mainly focuses on the diagnosis and active early warning of thermal faults of transformer bushings, thereby effectively solving the problem of intelligent diagnosis and early warning of thermal faults of main transformer bushings in current substations. The present invention effectively solves the problem of intelligent diagnosis and early warning of thermal faults of main transformer bushings in current substations, can plan the operation and maintenance schedule, make the power system intelligent and automated, and bring greater economic and social benefits to the safe and stable operation of the entire power system.

[0004] The existing transformer fault diagnosis and early warning methods have the following defects: The random fluctuations and noise contained in non-stationary data directly affect the model's learning and prediction capabilities. When the model attempts to learn patterns from this data, it will be disturbed by the noise, resulting in learning the noise rather than the true pattern, which in turn increases the model's prediction error. Models trained on datasets containing a lot of noise often have limited generalization capabilities. The model may overfit to the noise in the training data rather than learning the inherent structure of the data, resulting in poor performance on unseen data. If the weights are not adjusted according to the importance of the features, the model may overemphasize certain less important features and ignore features that have a greater impact on the prediction results, resulting in a decrease in prediction accuracy; models that do not consider feature weights may be more difficult to explain their decision-making process; unrestricted models may cause the weight coefficients to become very large or very small. Such extreme weight values ​​are not only difficult to explain but may also lead to numerical stability problems; the selection and adjustment of the model may be affected by the subjective judgment of the researchers, resulting in a lack of objectivity in the model and increased decision-making risks.

[0005] In view of this, the present invention proposes a transformer fault diagnosis and early warning method based on big data to solve the above problems. Summary of the Invention

[0006] In order to overcome the above-mentioned defects of the prior art and to achieve the above-mentioned purpose, the present invention provides the following technical solution: a transformer fault diagnosis and early warning method based on big data, comprising: S1. Collect transformer operating parameter data and environmental parameter data; S2. Preprocessing the operating parameter data and environmental parameter data of the transformer to obtain an operating parameter data set and an environmental parameter data set; performing feature extraction on the operating parameter data set and the environmental parameter data set to obtain a comprehensive feature data set; S3. Build a transformer fault warning model, input the comprehensive feature data set into the transformer fault warning model, and obtain a predicted transformer operating state; compare the predicted transformer operating state with a preset transformer operating state threshold to determine whether the transformer operating state is normal; S4. If the transformer is operating normally, the transformer operating status is continuously monitored; if the transformer is operating abnormally, the alarm mechanism is automatically triggered to generate fault warning information; transformer operating fault data is collected through the transformer monitoring terminal; S5. Build a transformer fault diagnosis model, input the transformer operation fault data into the transformer fault diagnosis model, and obtain the transformer fault type; S6. Take corresponding countermeasures according to the type of transformer fault; each module is connected through wired and / or wireless means.

[0007] Furthermore, the operating parameter data of the transformer includes electrical parameter data, mechanical parameter data, thermal parameter data and physical and chemical parameter data; the environmental parameter data includes ambient temperature data and ambient humidity data; The electrical parameter data includes current, voltage, power and load; the mechanical parameter data includes vibration signal and noise level; the thermal parameter data includes winding temperature and oil temperature; the physical and chemical parameter data includes dissolved gas analysis (DGA) in oil, oil quality parameters and insulation material status; The operating parameter data and environmental parameter data of the transformer are collected through the online monitoring system OLMS and sensors, and stored in the database management system DBMS.

[0008] Furthermore, the method of preprocessing the transformer operating parameter data and environmental parameter data to obtain the operating parameter data set and the environmental parameter data set includes: For the acquired operating parameter data and environmental parameter data, the linear interpolation method is used to fill missing values, and the cumulative distribution function is used to detect and remove outliers that significantly deviate from the overall trend of the data. The data after removing outliers are normalized and converted into a standard normal distribution with a mean of 0 and a standard deviation of 1 to eliminate the dimensional influence between the data, thus obtaining the preprocessed operating parameter data set and environmental parameter data set. The moving average method is used to stabilize the normalized data. The specific steps are as follows: S31. Preset a sliding window size m, where each window contains m data points; and adjust the sliding window size m using a sliding window adjustment formula, where the sliding window adjustment formula is: ;in, is the type of dataset; is the number of data points in the dataset; S32, initialize the sliding window, starting from the first data point of the current data set to the end, and create multiple windows according to the window size m; S33. Calculate the average value for each data point in the current window; S34, the sliding window moves forward one time step, moves the first data point out of the window, and includes the new data point in the window; S35, repeat steps S33 and S34 until the entire data set is traversed; S36. Summarize all the obtained average values ​​to obtain a smoothed data set. The specific mathematical formula is: ;in, For the time point The moving average at is the window size; For the data set The original value of the moment; Through the above steps, the preprocessed operating parameter data set and environmental parameter data set are obtained.

[0009] Furthermore, the method of extracting features from the operating parameter dataset and the environmental parameter dataset to obtain a comprehensive feature dataset includes: Fourier transform is used to extract frequency domain features for periodic analysis. The operating parameter data set and the environmental parameter data set are converted into spectrum form by applying the Fourier transform formula. The Fourier transform formula is: ;in, is the signal in the frequency domain; is the signal in the time domain; is the frequency; is an imaginary unit; is a complex exponential function; is the symbol for integral differential element; is the integral operation symbol; Wavelet transform is used to extract time-frequency features to capture the sudden changes and transient characteristics of the data; Morlet wavelet is selected as the wavelet mother function; continuous wavelet transform is applied to each time series signal of the operating parameter data and environmental parameter data to obtain the coefficients at different scales and time positions. ; The mathematical formula of wavelet transform is: ;in is the wavelet mother function; is the scale parameter; is the translation parameter; is the original signal to be analyzed; is the continuous wavelet transform system; is the normalization factor; The frequency domain features extracted by Fourier transform are integrated to obtain a new operation feature dataset, and the time-frequency features extracted by wavelet transform are integrated to obtain a new environmental feature dataset; the new operation feature dataset and the new environmental feature dataset are weighted fused to obtain a comprehensive feature dataset.

[0010] Furthermore, the method of weightedly fusing the new operation feature dataset and the new environment feature dataset to obtain a comprehensive feature dataset includes: S51, record the running feature data set as , the environmental feature dataset is recorded as , the weighted model is: ;in, It is a comprehensive feature dataset; is the weight coefficient of the running feature dataset; is the weight coefficient of the environmental feature dataset; S52. Design the weight coefficient for each feature data set in the weighted model: construct an objective function to minimize the error of the weighted model prediction; The objective function is: ; Among them, Objective function; is the number of samples in the comprehensive feature dataset; For the The predicted value of samples; For the The true value of the samples; S53. Introduction of Lagrange multipliers ,and ;in, To control the Lagrange multiplier function; constructing a Lagrangian function; the Lagrangian function is: ;in, is the Lagrangian function; S54. According to the Lagrangian function respectively 、 、 and Find the partial derivatives and set them equal to 0, and we get the following four equations: ; ; ; ; S55, through the Lagrange multiplier restriction model, control the Lagrange multiplier Function The Lagrange multiplier restriction model is: ;in, The restricted function ; is the number of samples in the comprehensive feature dataset.

[0011] Furthermore, the method for constructing the transformer fault early warning model includes: The dataset is divided into training set, validation set and test set. The dataset includes historical comprehensive feature dataset and corresponding transformer operating status. A transformer fault warning model is constructed. The transformer fault warning model includes input layer, hidden layer and output layer. The hidden layer uses ReLU activation function. The input layer of the model is a historical comprehensive feature dataset, and the number of neurons in the input layer matches the number of features in the historical comprehensive feature dataset. The output layer of the model is the transformer operating status, and the number of neurons in the output layer corresponds to the number of prediction targets, with one neuron outputting the predicted value. The transformer fault warning model is a fully connected neural network model. The mean square error is used as the loss function to measure the error between the model's predicted value and the actual value; the mean square error loss function is: ;in, is the number of datasets; Data points in the dataset The actual value of Data points in the dataset The predicted value of The transformer fault warning model is trained using the training set, and the model parameters are updated using the back-propagation algorithm to minimize the loss function. The performance of the transformer fault warning model is evaluated using the validation set by calculating the recall rate metric. The Adam optimization algorithm is selected as the optimizer, and the model is tuned according to the performance feedback of the validation set. The model parameters are adjusted until the performance no longer improves significantly or the preset number of iterations is reached. The performance of the model in the prediction task is evaluated using the test set, and the trained transformer fault warning model is used to predict the current comprehensive feature dataset to obtain the transformer operating status.

[0012] Furthermore, the method of comparing the predicted transformer operating state with a preset transformer operating state threshold to determine whether the transformer operating state is normal includes: If the predicted transformer operating state is greater than or equal to the preset transformer operating state threshold, the transformer operating state is determined to be normal; If the predicted transformer operating state is less than a preset transformer operating state threshold, the transformer operating state is determined to be abnormal.

[0013] Furthermore, the transformer operation fault data includes electrical parameter fault data, mechanical parameter fault data, thermal parameter fault data and physical and chemical parameter fault data.

[0014] Furthermore, the training method of the transformer fault diagnosis model includes: The dataset is divided into a training set, a test set, and a validation set. The dataset includes historical transformer operation fault data and the corresponding transformer fault types. A transformer fault diagnosis model is constructed. The transformer fault diagnosis model includes an input layer, a hidden layer, and an output layer. The input layer is historical transformer operation fault data, and the output layer is transformer fault types. The output layer is configured with neurons equal to the number of transformer fault types, with each neuron corresponding to the predicted probability of a defect type. The softmax function is used as the activation function. The transformer fault diagnosis model is a multi-layer perceptron (MLP) model. Multi-class cross entropy is used as the loss function of the model to measure the difference between the model's predicted value and the actual value; the multi-class cross entropy loss function is: ;in, is the average loss of the dataset; is the total number of samples in the dataset; is the number of transformer fault types; For the The samples for The true labels of each type; The model predicts The samples belong to The probability of each type; The transformer fault diagnosis model is trained using the training set and the model parameters are updated using the back-propagation algorithm to minimize the loss function. The performance of the transformer fault diagnosis model is evaluated using the validation set by calculating the accuracy index. The SGD optimization algorithm is selected as the optimizer, and the model is tuned according to the performance feedback of the validation set. The model parameters are adjusted until the performance no longer improves significantly or the preset stopping condition is reached. The performance of the model in the prediction task is evaluated using the test set, and the trained transformer fault diagnosis model is used to predict the current comprehensive feature data set to obtain the transformer fault type.

[0015] Furthermore, the method of taking corresponding countermeasures according to the transformer fault type includes: Replace aging or damp insulation materials, and repair or replace windings; eliminate multiple grounding points by using electric shock or adding resistors; filter or replace insulating oil to keep it clean; check equipment within the protection range, find and handle short-circuit faults; adjust load distribution to reduce the burden on the transformer, and add transformers if necessary; check short-circuit points, repair damaged parts, and strengthen short-circuit protection; check and tighten loose parts to eliminate abnormal vibration and noise.

[0016] The technical effects and advantages of the transformer fault diagnosis and early warning method based on big data of the present invention are as follows: The present invention stabilizes data through the moving average method, which can effectively reduce random fluctuations and noise in the data, making the data trend smoother, which is conducive to subsequent data analysis and modeling; dynamically adjusting the sliding window size according to the type of data set and the number of data points can more flexibly adapt to the characteristics of different data sets and improve the efficiency and accuracy of data processing; through data stabilization processing, the noise and fluctuations in the data can be reduced, making the model more stable and having better generalization ability.

[0017] By setting weight coefficients for the operational feature dataset and the environmental feature dataset separately, the model can flexibly adapt to changes in feature importance in different scenarios. This flexibility enables the model to more accurately reflect the actual situation and improve the accuracy and reliability of predictions. An objective function is constructed to minimize the error of the weighted model prediction, and the optimal weight coefficient is found by solving the partial derivatives. This ensures that the model's performance on a given dataset is optimal, that is, the difference between the predicted value and the true value is minimized.

[0018] Introducing Lagrange multipliers The corresponding Lagrange multiplier restriction model helps prevent model overfitting and improves the generalization ability of the model. At the same time, the Lagrange multiplier restriction model can also ensure that the weight coefficient is within a reasonable range and avoid the occurrence of extreme situations. It helps to reduce the interference of human factors and improve the objectivity and accuracy of the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 This is a flowchart of a transformer fault diagnosis and early warning method based on big data; Figure 2 This is a structural diagram of a transformer fault diagnosis and early warning system based on big data. DETAILED DESCRIPTION

[0020] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0021] Example 1 See also Figure 1 As shown, the transformer fault diagnosis and early warning method based on big data described in this embodiment includes: S1, collecting transformer operating parameter data and environmental parameter data; S2. Preprocessing the operating parameter data and environmental parameter data of the transformer to obtain an operating parameter data set and an environmental parameter data set; performing feature extraction on the operating parameter data set and the environmental parameter data set to obtain a comprehensive feature data set; S3. Build a transformer fault warning model, input the comprehensive feature data set into the transformer fault warning model, and obtain a predicted transformer operating state; compare the predicted transformer operating state with a preset transformer operating state threshold to determine whether the transformer operating state is normal; S4. If the transformer is operating normally, the transformer operating status is continuously monitored; if the transformer is operating abnormally, the alarm mechanism is automatically triggered to generate fault warning information; transformer operating fault data is collected through the transformer monitoring terminal; S5. Build a transformer fault diagnosis model, input the transformer operation fault data into the transformer fault diagnosis model, and obtain the transformer fault type; S6. Take corresponding countermeasures according to the type of transformer fault; each module is connected through wired and / or wireless means.

[0022] The operating parameter data of the transformer includes electrical parameter data, mechanical parameter data, thermal parameter data and physical and chemical parameter data; the environmental parameter data includes ambient temperature data and ambient humidity data; Electrical parameter data includes current, voltage, power and load; mechanical parameter data includes vibration signal and noise level; thermal parameter data includes winding temperature and oil temperature; physical and chemical parameter data includes dissolved gas analysis (DGA) in oil, oil quality parameters and insulation material status; The operating parameter data and environmental parameter data of the transformer are collected through the online monitoring system OLMS and sensors, and stored in the database management system DBMS.

[0023] The method of preprocessing the operating parameter data and the environmental parameter data of the transformer to obtain the operating parameter data set and the environmental parameter data set includes: For the acquired operating parameter data and environmental parameter data, the linear interpolation method is used to fill missing values, and the cumulative distribution function is used to detect and remove outliers that significantly deviate from the overall trend of the data. The data after removing outliers are normalized and converted into a standard normal distribution with a mean of 0 and a standard deviation of 1 to eliminate the dimensional influence between the data, thus obtaining the preprocessed operating parameter data set and environmental parameter data set. The moving average method is used to stabilize the normalized data. The specific steps are as follows: S31. Preset a sliding window size m, where each window contains m data points. The window size determines the number of data points considered each time the average value is calculated. Adjust the sliding window size m using a sliding window adjustment formula, which is: ;in, is the type of dataset; is the number of data points in the dataset; S32, initialize the sliding window, starting from the first data point of the current data set to the end, and create multiple windows according to the window size m; S33. Calculate the average value for each data point in the current window; S34, the sliding window moves forward one time step, moves the first data point out of the window, and includes the new data point in the window; S35, repeat steps S33 and S34 until the entire data set is traversed; S36. Summarize all the obtained average values ​​to obtain a smoothed data set. The specific mathematical formula is: ;in, For the time point The moving average at is the window size; For the data set The original value of the moment; Through the above steps, the preprocessed operating parameter data set and environmental parameter data set are obtained.

[0024] For example, suppose there is a normalized time series dataset representing the operating parameters of a device at different time points. The data is as follows: X=[2.1,2.5,2.3,2.8,3.0,2.7,2.9,3.2,3.5,3.3]; The number of data points, N, is 10. We use a moving average to smooth this data. Assume we set the sliding window size, m, to 3. This means that each time we calculate the average, we consider three data points in the current window. We'll use the sliding window adjustment formula to set m, but for this simple example, we'll just set it to 3. Starting from the first data point of the dataset, initialize the first sliding window, which contains the first m = 3 data points: window 1 = [2.1, 2.5, 2.3]; for window 1, calculate the mean ; Move the sliding window forward one time step, remove the first data point from the window, and include the new data point in the window: window 2 = [2.5, 2.3, 2.8]; Calculate the average value for window 2 ; Continue moving the window and calculating the average value of each window, summing up all the calculated moving averages to obtain the smoothed data set: Smoothed dataset = [2.3, 2.53, 2.7, 2.83, 2.87, 2.93, 3.2, 3.33]; Methods for extracting features from the operating parameter dataset and the environmental parameter dataset to obtain a comprehensive feature dataset include: Fourier transform is used to extract frequency domain features for periodic analysis. The operating parameter data set and the environmental parameter data set are converted into spectrum form by applying the Fourier transform formula. The Fourier transform formula is: ;in, is the signal in the frequency domain; is the signal in the time domain; is the frequency; is an imaginary unit, usually ; is a complex exponential function; is the symbol for integral differential element; is an integral operation symbol; the frequency domain characteristics include the main frequency components, the amplitude of each frequency component, the phase of each frequency component and the frequency domain statistical characteristics; Wavelet transform is used to extract time-frequency features to capture the sudden changes and transient characteristics of the data; Morlet wavelet is selected as the wavelet mother function; continuous wavelet transform is applied to each time series signal of the operating parameter data and environmental parameter data to obtain the coefficients at different scales and time positions. ; The mathematical formula of wavelet transform is: ;in is the wavelet mother function; is a scale parameter used to control the scaling of the wavelet. A positive scale factor indicates that the wavelet is enlarged or reduced. A large scale indicates that low-frequency information (long time interval) is analyzed, and a small scale indicates that high-frequency information (short time interval) is analyzed. is the translation parameter, which is used to control the movement of the wavelet in time. , which can perform localized analysis on different locations of the signal; is the original signal to be analyzed; is the continuous wavelet transform system, which is the original signal The transformation results at different scales and displacements represent the signal components in scale and location; is a normalization factor used to keep the energy of the wavelet unchanged at different scales, ensuring the energy consistency of the wavelet transform at different scales and making the analysis results comparable; The frequency domain features extracted by Fourier transform are integrated to obtain a new operation feature dataset, and the time-frequency features extracted by wavelet transform are integrated to obtain a new environmental feature dataset; the new operation feature dataset and the new environmental feature dataset are weighted fused to obtain a comprehensive feature dataset.

[0025] Methods for performing weighted fusion of the new operation feature dataset and the new environment feature dataset to obtain a comprehensive feature dataset include: S51, record the running feature data set as , the environmental feature dataset is recorded as , the weighted model is: ;in, It is a comprehensive feature dataset; is the weight coefficient of the running feature dataset; is the weight coefficient of the environmental feature dataset; For example, the weight coefficients of the feature dataset are run The weight coefficient of the environmental feature dataset is 0.6. is 0.4; S52. Design the weight coefficient for each feature data set in the weighted model: construct an objective function to minimize the error of the weighted model prediction; The objective function is: ; Among them, Objective function; is the number of samples in the comprehensive feature dataset; For the The predicted value of samples; For the The true value of the samples; S53. Introduction of Lagrange multipliers ,and ;in, To control the Lagrange multiplier function; constructing a Lagrangian function; the Lagrangian function is: ;in, is the Lagrangian function; S54. According to the Lagrangian function respectively 、 、 and Find the partial derivatives and set them equal to 0, and we get the following four equations: ; ; ; ; S55, through the Lagrange multiplier restriction model, control the Lagrange multiplier Function The Lagrange multiplier restriction model is: ;in, The restricted function ; is the number of samples in the comprehensive feature dataset.

[0026] For example, the number of samples in the synthetic feature dataset The weight coefficient of the running feature dataset is 10 The weight coefficient of the environmental feature dataset is 0.6. is 0.4; then the restricted function The value is approximately 0.92.

[0027] The construction method of the transformer fault early warning model includes: The dataset is divided into training set, validation set and test set. The dataset includes historical comprehensive feature dataset and corresponding transformer operating status. A transformer fault warning model is constructed. The transformer fault warning model includes input layer, hidden layer and output layer. The hidden layer uses ReLU activation function. The input layer of the model is a historical comprehensive feature dataset, and the number of neurons in the input layer matches the number of features in the historical comprehensive feature dataset. The output layer of the model is the transformer operating status, and the number of neurons in the output layer corresponds to the number of prediction targets, with one neuron outputting the predicted value. The transformer fault warning model is a fully connected neural network model. The mean square error is used as the loss function to measure the error between the model's predicted value and the actual value; the mean square error loss function is: ;in, is the number of datasets; Data points in the dataset The actual value of Data points in the dataset The predicted value of The transformer fault warning model is trained using the training set, and the model parameters are updated using the back-propagation algorithm to minimize the loss function. The performance of the transformer fault warning model is evaluated using the validation set by calculating the recall rate metric. The Adam optimization algorithm is selected as the optimizer, and the model is tuned according to the performance feedback of the validation set. The model parameters are adjusted until the performance no longer improves significantly or the preset number of iterations is reached. The performance of the model in the prediction task is evaluated using the test set, and the trained transformer fault warning model is used to predict the current comprehensive feature dataset to obtain the transformer operating status.

[0028] The method of comparing the predicted transformer operating state with a preset transformer operating state threshold to determine whether the transformer operating state is normal includes: If the predicted transformer operating state is greater than or equal to the preset transformer operating state threshold, the transformer operating state is determined to be normal; If the predicted transformer operating state is less than a preset transformer operating state threshold, the transformer operating state is determined to be abnormal.

[0029] Transformer operation fault data includes electrical parameter fault data, mechanical parameter fault data, thermal parameter fault data and physical and chemical parameter fault data.

[0030] The training method of the transformer fault diagnosis model includes: The dataset is divided into a training set, a test set, and a validation set. The dataset includes historical transformer operation fault data and the corresponding transformer fault types. A transformer fault diagnosis model is constructed. The transformer fault diagnosis model includes an input layer, a hidden layer, and an output layer. The input layer is historical transformer operation fault data, and the output layer is transformer fault types. The output layer is configured with neurons equal to the number of transformer fault types, with each neuron corresponding to the predicted probability of a defect type. The softmax function is used as the activation function. The transformer fault diagnosis model is a multi-layer perceptron (MLP) model. Multi-class cross entropy is used as the loss function of the model to measure the difference between the model's predicted value and the actual value; the multi-class cross entropy loss function is: ;in, is the average loss of the dataset; is the total number of samples in the dataset; is the number of transformer fault types; For the The samples for The true labels of each type; The model predicts The samples belong to The probability of each type; The transformer fault diagnosis model is trained using the training set and the model parameters are updated using the back-propagation algorithm to minimize the loss function. The performance of the transformer fault diagnosis model is evaluated using the validation set by calculating the accuracy index. The SGD optimization algorithm is selected as the optimizer, and the model is tuned according to the performance feedback of the validation set. The model parameters are adjusted until the performance no longer improves significantly or the preset stopping condition is reached. The performance of the model in the prediction task is evaluated using the test set, and the trained transformer fault diagnosis model is used to predict the current comprehensive feature data set to obtain the transformer fault type.

[0031] Methods for taking appropriate countermeasures based on the type of transformer fault include: Replace aging or damp insulation materials, and repair or replace windings; eliminate multiple grounding points by using electric shock or adding resistors; filter or replace insulating oil to keep it clean; check equipment within the protection range, find and handle short-circuit faults; adjust load distribution to reduce the burden on the transformer, and add transformers if necessary; check short-circuit points, repair damaged parts, and strengthen short-circuit protection; check and tighten loose parts to eliminate abnormal vibration and noise.

[0032] The preset transformer operating state threshold is set by the staff, and different transformer operating states are collected through the transformer monitoring terminal, and the average value of multiple transformer operating states is taken as the preset transformer operating state threshold.

[0033] This embodiment uses the moving average method to smooth the data, which can effectively reduce random fluctuations and noise in the data, make the data trend smoother, and facilitate subsequent data analysis and modeling; dynamically adjust the sliding window size according to the type of data set and the number of data points, which can more flexibly adapt to the characteristics of different data sets and improve the efficiency and accuracy of data processing; through data smoothing processing, the noise and fluctuations in the data can be reduced, making the model more stable and having better generalization ability.

[0034] By setting weight coefficients for the operational feature dataset and the environmental feature dataset separately, the model can flexibly adapt to changes in feature importance in different scenarios. This flexibility enables the model to more accurately reflect the actual situation and improve the accuracy and reliability of predictions. An objective function is constructed to minimize the error of the weighted model prediction, and the optimal weight coefficient is found by solving the partial derivatives. This ensures that the model's performance on a given dataset is optimal, that is, the difference between the predicted value and the true value is minimized.

[0035] Introducing Lagrange multipliers The corresponding Lagrange multiplier restriction model helps prevent model overfitting and improves the generalization ability of the model. At the same time, the Lagrange multiplier restriction model can also ensure that the weight coefficient is within a reasonable range and avoid the occurrence of extreme situations. It helps to reduce the interference of human factors and improve the objectivity and accuracy of the model.

[0036] Example 2 See also Figure 2 As shown, for the parts not described in detail in this embodiment, please refer to the description of Example 1. A transformer fault diagnosis and early warning system based on big data is provided, including: Data acquisition module, used to collect transformer operating parameter data and environmental parameter data; The data processing module is used to pre-process the operating parameter data and environmental parameter data of the transformer to obtain an operating parameter data set and an environmental parameter data set; and perform feature extraction on the operating parameter data set and the environmental parameter data set to obtain a comprehensive feature data set; The fault warning module is used to build a transformer fault warning model, input the comprehensive feature data set into the transformer fault warning model, and obtain the predicted transformer operating status; the predicted transformer operating status is compared with the preset transformer operating status threshold to determine whether the transformer operating status is normal; Automatic alarm module: If the transformer is operating normally, it will continuously monitor the transformer's operating status; if the transformer is operating abnormally, it will automatically trigger the alarm mechanism and generate fault warning information; the transformer operating fault data will be collected through the transformer monitoring terminal; The fault diagnosis module is used to build a transformer fault diagnosis model, input the transformer operation fault data into the transformer fault diagnosis model, and obtain the transformer fault type; The fault response module is used to take corresponding response measures according to the type of transformer fault.

[0037] Since the electronic device described in this embodiment is an electronic device used to implement a transformer fault diagnosis and early warning method based on big data in the embodiment of this application, based on the transformer fault diagnosis and early warning method based on big data described in the embodiment of this application, those skilled in the art will be able to understand the specific implementation of the electronic device of this embodiment and its various variations, so how the electronic device implements the method in the embodiment of this application will not be described in detail here. As long as those skilled in the art implement the electronic device used in the transformer fault diagnosis and early warning method based on big data in the embodiment of this application, it falls within the scope of protection of this application.

[0038] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters and thresholds in the formulas are set by technicians in this field according to actual conditions.

[0039] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiment. All technical solutions based on the principles of the present invention are within the scope of protection of the present invention. It should be noted that for users of ordinary skill in the art, various improvements and modifications that do not depart from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A transformer fault diagnosis and early warning method based on big data, characterized in that: include: S1. Collect transformer operating parameter data and environmental parameter data; S2. Preprocessing the operating parameter data and environmental parameter data of the transformer to obtain an operating parameter data set and an environmental parameter data set; performing feature extraction on the operating parameter data set and the environmental parameter data set to obtain a comprehensive feature data set; S3. Build a transformer fault warning model, input the comprehensive feature data set into the transformer fault warning model, and obtain a predicted transformer operating state; compare the predicted transformer operating state with a preset transformer operating state threshold to determine whether the transformer operating state is normal; S4. If the transformer is operating normally, continue to monitor the transformer operating status; If the transformer is operating abnormally, the alarm mechanism will be automatically triggered to generate fault warning information; transformer operation fault data will be collected through the transformer monitoring terminal; S5. Build a transformer fault diagnosis model, input the transformer operation fault data into the transformer fault diagnosis model, and obtain the transformer fault type; S6. Take corresponding countermeasures according to the type of transformer fault; each module is connected through wired and / or wireless means.

2. A transformer fault diagnosis and early warning method based on big data according to claim 1, characterized in that: The operating parameter data of the transformer includes electrical parameter data, mechanical parameter data, thermal parameter data and physical and chemical parameter data; the environmental parameter data includes ambient temperature data and ambient humidity data; The electrical parameter data includes current, voltage, power and load; the mechanical parameter data includes vibration signal and noise level; the thermal parameter data includes winding temperature and oil temperature; the physical and chemical parameter data includes dissolved gas analysis (DGA) in oil, oil quality parameters and insulation material status; The operating parameter data and environmental parameter data of the transformer are collected through the online monitoring system OLMS and sensors, and stored in the database management system DBMS.

3. A transformer fault diagnosis and early warning method based on big data according to claim 2, characterized in that: The method of preprocessing the transformer operating parameter data and environmental parameter data to obtain the operating parameter data set and the environmental parameter data set includes: For the acquired operating parameter data and environmental parameter data, the linear interpolation method is used to fill missing values, and the cumulative distribution function is used to detect and remove outliers that significantly deviate from the overall trend of the data. The data after removing outliers are normalized and converted into a standard normal distribution with a mean of 0 and a standard deviation of 1 to eliminate the dimensional influence between the data, thus obtaining the preprocessed operating parameter data set and environmental parameter data set. The moving average method is used to stabilize the normalized data. The specific steps are as follows: S31. Preset a sliding window size m, where each window contains m data points; and adjust the sliding window size m using a sliding window adjustment formula, where the sliding window adjustment formula is: ;in, is the type of dataset; is the number of data points in the dataset; S32, initialize the sliding window, starting from the first data point of the current data set to the end, and create multiple windows according to the window size m; S33. Calculate the average value for each data point in the current window; S34, the sliding window moves forward one time step, moves the first data point out of the window, and includes the new data point in the window; S35, repeat steps S33 and S34 until the entire data set is traversed; S36. Summarize all the obtained average values ​​to obtain a smoothed data set. The specific mathematical formula is: ;in, For the time point The moving average at is the window size; For the data set The original value of the moment; Through the above steps, the preprocessed operating parameter data set and environmental parameter data set are obtained.

4. A transformer fault diagnosis and early warning method based on big data according to claim 3, characterized in that: The method of extracting features from the operating parameter dataset and the environmental parameter dataset to obtain a comprehensive feature dataset includes: Fourier transform is used to extract frequency domain features for periodic analysis. The operating parameter data set and the environmental parameter data set are converted into spectrum form by applying the Fourier transform formula. The Fourier transform formula is: ;in, is the signal in the frequency domain; is the signal in the time domain; is the frequency; is an imaginary unit; is a complex exponential function; is the symbol for integral differential element; is the integral operation symbol; Wavelet transform is used to extract time-frequency features to capture the sudden changes and transient characteristics of the data; Morlet wavelet is selected as the wavelet mother function; continuous wavelet transform is applied to each time series signal of the operating parameter data and environmental parameter data to obtain the coefficients at different scales and time positions. ; The mathematical formula of wavelet transform is: ;in is the wavelet mother function; is the scale parameter; is the translation parameter; is the original signal to be analyzed; is the continuous wavelet transform system; is the normalization factor; The frequency domain features extracted by Fourier transform are integrated to obtain a new operation feature dataset, and the time-frequency features extracted by wavelet transform are integrated to obtain a new environmental feature dataset; the new operation feature dataset and the new environmental feature dataset are weighted fused to obtain a comprehensive feature dataset.

5. A transformer fault diagnosis and early warning method based on big data according to claim 4, characterized in that: The method of weightedly fusing the new operation feature dataset and the new environment feature dataset to obtain a comprehensive feature dataset includes: S51, record the running feature data set as , the environmental feature dataset is recorded as , the weighted model is: ;in, It is a comprehensive feature dataset; is the weight coefficient of the running feature dataset; is the weight coefficient of the environmental feature dataset; S52. Design the weight coefficient for each feature data set in the weighted model: construct an objective function to minimize the error of the weighted model prediction; The objective function is: ; Among them, Objective function; is the number of samples in the comprehensive feature dataset; For the The predicted value of samples; For the The true value of the samples; S53. Introduction of Lagrange multipliers ,and ;in, To control the Lagrange multiplier function; constructing a Lagrangian function; the Lagrangian function is: ;in, is the Lagrangian function; S54, according to the Lagrangian function respectively 、 、 and Find the partial derivatives and set them equal to 0, and we get the following four equations: ； ； ； ； S55, through the Lagrange multiplier restriction model, control the Lagrange multiplier Function The Lagrange multiplier restriction model is: ;in, The restricted function ; is the number of samples in the comprehensive feature dataset.

6. A transformer fault diagnosis and early warning method based on big data according to claim 5, characterized in that: The method for constructing the transformer fault early warning model includes: The dataset is divided into training set, validation set and test set. The dataset includes historical comprehensive feature dataset and corresponding transformer operating status. A transformer fault warning model is constructed. The transformer fault warning model includes input layer, hidden layer and output layer. The hidden layer uses ReLU activation function. The input layer of the model is a historical comprehensive feature dataset, and the number of neurons in the input layer matches the number of features in the historical comprehensive feature dataset. The output layer of the model is the transformer operating status, and the number of neurons in the output layer corresponds to the number of prediction targets, with one neuron outputting the predicted value. The transformer fault warning model is a fully connected neural network model. The mean square error is used as the loss function to measure the error between the model's predicted value and the actual value; the mean square error loss function is: ;in, is the number of datasets; Data points in the dataset The actual value of Data points in the dataset The predicted value of The transformer fault warning model is trained using the training set, and the model parameters are updated using the back-propagation algorithm to minimize the loss function. The performance of the transformer fault warning model is evaluated using the validation set by calculating the recall rate metric. The Adam optimization algorithm is selected as the optimizer, and the model is tuned according to the performance feedback of the validation set. The model parameters are adjusted until the performance no longer improves significantly or the preset number of iterations is reached. The performance of the model in the prediction task is evaluated using the test set, and the trained transformer fault warning model is used to predict the current comprehensive feature dataset to obtain the transformer operating status.

7. A transformer fault diagnosis and early warning method based on big data according to claim 6, characterized in that: The method of comparing the predicted transformer operating state with a preset transformer operating state threshold to determine whether the transformer operating state is normal includes: If the predicted transformer operating state is greater than or equal to the preset transformer operating state threshold, the transformer operating state is determined to be normal; If the predicted transformer operating state is less than a preset transformer operating state threshold, the transformer operating state is determined to be abnormal.

8. The transformer fault diagnosis and early warning method based on big data according to claim 7 is characterized in that: The transformer operation fault data includes electrical parameter fault data, mechanical parameter fault data, thermal parameter fault data and physical and chemical parameter fault data.

9. The transformer fault diagnosis and early warning method based on big data according to claim 8 is characterized in that: The training method of the transformer fault diagnosis model includes: The dataset is divided into a training set, a test set, and a validation set. The dataset includes historical transformer operation fault data and the corresponding transformer fault types. A transformer fault diagnosis model is constructed. The transformer fault diagnosis model includes an input layer, a hidden layer, and an output layer. The input layer is historical transformer operation fault data, and the output layer is transformer fault types. The output layer is configured with neurons equal to the number of transformer fault types, with each neuron corresponding to the predicted probability of a defect type. The softmax function is used as the activation function. The transformer fault diagnosis model is a multi-layer perceptron (MLP) model. Multi-class cross entropy is used as the loss function of the model to measure the difference between the model's predicted value and the actual value; the multi-class cross entropy loss function is: ;in, is the average loss of the dataset; is the total number of samples in the dataset; is the number of transformer fault types; For the The samples for The true labels of each type; The model predicts The samples belong to The probability of each type; The transformer fault diagnosis model is trained using the training set and the model parameters are updated using the back-propagation algorithm to minimize the loss function. The performance of the transformer fault diagnosis model is evaluated using the validation set by calculating the accuracy index. The SGD optimization algorithm is selected as the optimizer, and the model is tuned according to the performance feedback of the validation set. The model parameters are adjusted until the performance no longer improves significantly or the preset stopping condition is reached. The performance of the model in the prediction task is evaluated using the test set, and the trained transformer fault diagnosis model is used to predict the current comprehensive feature data set to obtain the transformer fault type.

10. A transformer fault diagnosis and early warning method based on big data according to claim 9, characterized in that: The method of taking corresponding countermeasures according to the transformer fault type includes: Replace aging or damp insulation materials, and repair or replace windings; eliminate multiple grounding points by using electric shock or adding resistors; filter or replace insulating oil to keep it clean; check equipment within the protection range, find and handle short-circuit faults; adjust load distribution to reduce the burden on the transformer, and add transformers if necessary; check short-circuit points, repair damaged parts, and strengthen short-circuit protection; check and tighten loose parts to eliminate abnormal vibration and noise.

Images